Importing some important libraries required for Telco-Customer-Churn Data Analysis¶

Here, the dataset is downloaded from https://www.kaggle.com/datasets/blastchar/telco-customer-churn?resource=download

This dataset consists of 7043 rows(records) and 21 columns(features). There are 20 independent features and 1 dependent feature("Churn"). The problem statement is, why the customers are moving out(Churn is nothing but leaving out of the business) of the business. We have to give solution to the stakeholder, to solve this problem and to provide a model which predicts customer churn.
In [1]:
import pandas as pd
import numpy as np
import tensorflow as tf
from sklearn import preprocessing
from sklearn.model_selection import train_test_split
In [2]:
# reading the csv file from the current directory
df = pd.read_csv('WA_Fn-UseC_-Telco-Customer-Churn.csv')
In [3]:
# printing the df variable which holds churn dataset from IBM
df
Out[3]:
customerID gender SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity ... DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn
0 7590-VHVEG Female 0 Yes No 1 No No phone service DSL No ... No No No No Month-to-month Yes Electronic check 29.85 29.85 No
1 5575-GNVDE Male 0 No No 34 Yes No DSL Yes ... Yes No No No One year No Mailed check 56.95 1889.5 No
2 3668-QPYBK Male 0 No No 2 Yes No DSL Yes ... No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes
3 7795-CFOCW Male 0 No No 45 No No phone service DSL Yes ... Yes Yes No No One year No Bank transfer (automatic) 42.30 1840.75 No
4 9237-HQITU Female 0 No No 2 Yes No Fiber optic No ... No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 6840-RESVB Male 0 Yes Yes 24 Yes Yes DSL Yes ... Yes Yes Yes Yes One year Yes Mailed check 84.80 1990.5 No
7039 2234-XADUH Female 0 Yes Yes 72 Yes Yes Fiber optic No ... Yes No Yes Yes One year Yes Credit card (automatic) 103.20 7362.9 No
7040 4801-JZAZL Female 0 Yes Yes 11 No No phone service DSL Yes ... No No No No Month-to-month Yes Electronic check 29.60 346.45 No
7041 8361-LTMKD Male 1 Yes No 4 Yes Yes Fiber optic No ... No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes
7042 3186-AJIEK Male 0 No No 66 Yes No Fiber optic Yes ... Yes Yes Yes Yes Two year Yes Bank transfer (automatic) 105.65 6844.5 No

7043 rows × 21 columns

In [4]:
# describe gives a small statistical summary about the dataset(df)
df.describe()
Out[4]:
SeniorCitizen tenure MonthlyCharges
count 7043.000000 7043.000000 7043.000000
mean 0.162147 32.371149 64.761692
std 0.368612 24.559481 30.090047
min 0.000000 0.000000 18.250000
25% 0.000000 9.000000 35.500000
50% 0.000000 29.000000 70.350000
75% 0.000000 55.000000 89.850000
max 1.000000 72.000000 118.750000
In [5]:
# dtypes gives the datatypes of the individual features in the dataframe
df.dtypes
Out[5]:
customerID           object
gender               object
SeniorCitizen         int64
Partner              object
Dependents           object
tenure                int64
PhoneService         object
MultipleLines        object
InternetService      object
OnlineSecurity       object
OnlineBackup         object
DeviceProtection     object
TechSupport          object
StreamingTV          object
StreamingMovies      object
Contract             object
PaperlessBilling     object
PaymentMethod        object
MonthlyCharges      float64
TotalCharges         object
Churn                object
dtype: object

Only few of the features having the datatype of Integer/Float. We have remove some unwanted features and to perform some Feature Engineering task.¶

Feature Engineering¶

1. The First feature is, "customerID". Definitely 'customerID' is not the reason for churn. So, we are removing the feature 'customerID'¶

In [6]:
df = df.drop(['customerID'], axis = 1)
In [7]:
df
Out[7]:
gender SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn
0 Female 0 Yes No 1 No No phone service DSL No Yes No No No No Month-to-month Yes Electronic check 29.85 29.85 No
1 Male 0 No No 34 Yes No DSL Yes No Yes No No No One year No Mailed check 56.95 1889.5 No
2 Male 0 No No 2 Yes No DSL Yes Yes No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes
3 Male 0 No No 45 No No phone service DSL Yes No Yes Yes No No One year No Bank transfer (automatic) 42.30 1840.75 No
4 Female 0 No No 2 Yes No Fiber optic No No No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 Male 0 Yes Yes 24 Yes Yes DSL Yes No Yes Yes Yes Yes One year Yes Mailed check 84.80 1990.5 No
7039 Female 0 Yes Yes 72 Yes Yes Fiber optic No Yes Yes No Yes Yes One year Yes Credit card (automatic) 103.20 7362.9 No
7040 Female 0 Yes Yes 11 No No phone service DSL Yes No No No No No Month-to-month Yes Electronic check 29.60 346.45 No
7041 Male 1 Yes No 4 Yes Yes Fiber optic No No No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes
7042 Male 0 No No 66 Yes No Fiber optic Yes No Yes Yes Yes Yes Two year Yes Bank transfer (automatic) 105.65 6844.5 No

7043 rows × 20 columns

2. The second feature is 'gender'. Let's look at the feature indepth.¶

In [8]:
df['gender'].isnull().sum()
Out[8]:
0

There is no null value in the 'gender' feature.

In [9]:
df['gender'].unique()
Out[9]:
array(['Female', 'Male'], dtype=object)

There is a two unique values in the 'gender' feature(1. Female, 2. Male). Let's do one hot encoding for this feature using get_dummies in pandas.

In [10]:
gender = pd.get_dummies(df['gender'], prefix = 'gender', dtype = 'int', drop_first = True)
In [11]:
gender
Out[11]:
gender_Male
0 0
1 1
2 1
3 1
4 0
... ...
7038 1
7039 0
7040 0
7041 1
7042 1

7043 rows × 1 columns

Removing the gender column from original dataframe.

In [12]:
df = df.drop(['gender'], axis = 1)
In [13]:
df
Out[13]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn
0 0 Yes No 1 No No phone service DSL No Yes No No No No Month-to-month Yes Electronic check 29.85 29.85 No
1 0 No No 34 Yes No DSL Yes No Yes No No No One year No Mailed check 56.95 1889.5 No
2 0 No No 2 Yes No DSL Yes Yes No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes
3 0 No No 45 No No phone service DSL Yes No Yes Yes No No One year No Bank transfer (automatic) 42.30 1840.75 No
4 0 No No 2 Yes No Fiber optic No No No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 Yes Yes 24 Yes Yes DSL Yes No Yes Yes Yes Yes One year Yes Mailed check 84.80 1990.5 No
7039 0 Yes Yes 72 Yes Yes Fiber optic No Yes Yes No Yes Yes One year Yes Credit card (automatic) 103.20 7362.9 No
7040 0 Yes Yes 11 No No phone service DSL Yes No No No No No Month-to-month Yes Electronic check 29.60 346.45 No
7041 1 Yes No 4 Yes Yes Fiber optic No No No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes
7042 0 No No 66 Yes No Fiber optic Yes No Yes Yes Yes Yes Two year Yes Bank transfer (automatic) 105.65 6844.5 No

7043 rows × 19 columns

Now, we use concat from pandas to join 'df' and 'gender'.

In [14]:
df = pd.concat((df, gender), axis = 1)
In [15]:
df
Out[15]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male
0 0 Yes No 1 No No phone service DSL No Yes No No No No Month-to-month Yes Electronic check 29.85 29.85 No 0
1 0 No No 34 Yes No DSL Yes No Yes No No No One year No Mailed check 56.95 1889.5 No 1
2 0 No No 2 Yes No DSL Yes Yes No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes 1
3 0 No No 45 No No phone service DSL Yes No Yes Yes No No One year No Bank transfer (automatic) 42.30 1840.75 No 1
4 0 No No 2 Yes No Fiber optic No No No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 Yes Yes 24 Yes Yes DSL Yes No Yes Yes Yes Yes One year Yes Mailed check 84.80 1990.5 No 1
7039 0 Yes Yes 72 Yes Yes Fiber optic No Yes Yes No Yes Yes One year Yes Credit card (automatic) 103.20 7362.9 No 0
7040 0 Yes Yes 11 No No phone service DSL Yes No No No No No Month-to-month Yes Electronic check 29.60 346.45 No 0
7041 1 Yes No 4 Yes Yes Fiber optic No No No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes 1
7042 0 No No 66 Yes No Fiber optic Yes No Yes Yes Yes Yes Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1

7043 rows × 20 columns

3. Next feature is partner. Let's do feature engineering on it.¶

In [16]:
df['Partner'].isnull().sum()
Out[16]:
0
In [17]:
df['Partner'].unique()
Out[17]:
array(['Yes', 'No'], dtype=object)

There are only two unique values('Yes' and 'No'). Let's do a label encoding. It means that for 'Yes', we have replace it with 1 and for 'No' we have to replace it with 0.

In [18]:
le = preprocessing.LabelEncoder()
In [19]:
df['Partner'] = le.fit_transform(df['Partner'])
In [20]:
df
Out[20]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male
0 0 1 No 1 No No phone service DSL No Yes No No No No Month-to-month Yes Electronic check 29.85 29.85 No 0
1 0 0 No 34 Yes No DSL Yes No Yes No No No One year No Mailed check 56.95 1889.5 No 1
2 0 0 No 2 Yes No DSL Yes Yes No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes 1
3 0 0 No 45 No No phone service DSL Yes No Yes Yes No No One year No Bank transfer (automatic) 42.30 1840.75 No 1
4 0 0 No 2 Yes No Fiber optic No No No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 Yes 24 Yes Yes DSL Yes No Yes Yes Yes Yes One year Yes Mailed check 84.80 1990.5 No 1
7039 0 1 Yes 72 Yes Yes Fiber optic No Yes Yes No Yes Yes One year Yes Credit card (automatic) 103.20 7362.9 No 0
7040 0 1 Yes 11 No No phone service DSL Yes No No No No No Month-to-month Yes Electronic check 29.60 346.45 No 0
7041 1 1 No 4 Yes Yes Fiber optic No No No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes 1
7042 0 0 No 66 Yes No Fiber optic Yes No Yes Yes Yes Yes Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1

7043 rows × 20 columns

4. 'Dependet' feature¶

In [21]:
df['Dependents'].isnull().sum()
Out[21]:
0
In [22]:
df['Dependents'].unique()
Out[22]:
array(['No', 'Yes'], dtype=object)

There are only two unique values('Yes' and 'No'). Let's do a label encoding. It means that for 'Yes', we have replace it with 1 and for 'No' we have to replace it with 0.

In [23]:
df['Dependents'] = le.fit_transform(df['Dependents'])
In [24]:
df
Out[24]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male
0 0 1 0 1 No No phone service DSL No Yes No No No No Month-to-month Yes Electronic check 29.85 29.85 No 0
1 0 0 0 34 Yes No DSL Yes No Yes No No No One year No Mailed check 56.95 1889.5 No 1
2 0 0 0 2 Yes No DSL Yes Yes No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes 1
3 0 0 0 45 No No phone service DSL Yes No Yes Yes No No One year No Bank transfer (automatic) 42.30 1840.75 No 1
4 0 0 0 2 Yes No Fiber optic No No No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 Yes Yes DSL Yes No Yes Yes Yes Yes One year Yes Mailed check 84.80 1990.5 No 1
7039 0 1 1 72 Yes Yes Fiber optic No Yes Yes No Yes Yes One year Yes Credit card (automatic) 103.20 7362.9 No 0
7040 0 1 1 11 No No phone service DSL Yes No No No No No Month-to-month Yes Electronic check 29.60 346.45 No 0
7041 1 1 0 4 Yes Yes Fiber optic No No No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes 1
7042 0 0 0 66 Yes No Fiber optic Yes No Yes Yes Yes Yes Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1

7043 rows × 20 columns

5.PhoneService¶

In [25]:
df['PhoneService'].isnull().sum()
Out[25]:
0
In [26]:
df['PhoneService'].unique()
Out[26]:
array(['No', 'Yes'], dtype=object)

There are only two unique values('Yes' and 'No'). Let's do a label encoding. It means that for 'Yes', we have replace it with 1 and for 'No' we have to replace it with 0.

In [27]:
df['PhoneService'] = le.fit_transform(df['PhoneService'])
In [28]:
df
Out[28]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male
0 0 1 0 1 0 No phone service DSL No Yes No No No No Month-to-month Yes Electronic check 29.85 29.85 No 0
1 0 0 0 34 1 No DSL Yes No Yes No No No One year No Mailed check 56.95 1889.5 No 1
2 0 0 0 2 1 No DSL Yes Yes No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes 1
3 0 0 0 45 0 No phone service DSL Yes No Yes Yes No No One year No Bank transfer (automatic) 42.30 1840.75 No 1
4 0 0 0 2 1 No Fiber optic No No No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 Yes DSL Yes No Yes Yes Yes Yes One year Yes Mailed check 84.80 1990.5 No 1
7039 0 1 1 72 1 Yes Fiber optic No Yes Yes No Yes Yes One year Yes Credit card (automatic) 103.20 7362.9 No 0
7040 0 1 1 11 0 No phone service DSL Yes No No No No No Month-to-month Yes Electronic check 29.60 346.45 No 0
7041 1 1 0 4 1 Yes Fiber optic No No No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes 1
7042 0 0 0 66 1 No Fiber optic Yes No Yes Yes Yes Yes Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1

7043 rows × 20 columns

6. MultipleLines¶

In [29]:
df['MultipleLines'].isnull().sum()
Out[29]:
0
In [30]:
df['MultipleLines'].unique()
Out[30]:
array(['No phone service', 'No', 'Yes'], dtype=object)
In [31]:
df.loc[df['MultipleLines'] == "No phone service", "MultipleLines"] = "No"
In [32]:
df
Out[32]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male
0 0 1 0 1 0 No DSL No Yes No No No No Month-to-month Yes Electronic check 29.85 29.85 No 0
1 0 0 0 34 1 No DSL Yes No Yes No No No One year No Mailed check 56.95 1889.5 No 1
2 0 0 0 2 1 No DSL Yes Yes No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes 1
3 0 0 0 45 0 No DSL Yes No Yes Yes No No One year No Bank transfer (automatic) 42.30 1840.75 No 1
4 0 0 0 2 1 No Fiber optic No No No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 Yes DSL Yes No Yes Yes Yes Yes One year Yes Mailed check 84.80 1990.5 No 1
7039 0 1 1 72 1 Yes Fiber optic No Yes Yes No Yes Yes One year Yes Credit card (automatic) 103.20 7362.9 No 0
7040 0 1 1 11 0 No DSL Yes No No No No No Month-to-month Yes Electronic check 29.60 346.45 No 0
7041 1 1 0 4 1 Yes Fiber optic No No No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes 1
7042 0 0 0 66 1 No Fiber optic Yes No Yes Yes Yes Yes Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1

7043 rows × 20 columns

In [33]:
df['MultipleLines'].unique()
Out[33]:
array(['No', 'Yes'], dtype=object)

We change that 'No phone service' to 'No'. Now, we can perform Label Encoding using the object 'le'.

In [34]:
df['MultipleLines'] = le.fit_transform(df['MultipleLines'])
In [35]:
df
Out[35]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male
0 0 1 0 1 0 0 DSL No Yes No No No No Month-to-month Yes Electronic check 29.85 29.85 No 0
1 0 0 0 34 1 0 DSL Yes No Yes No No No One year No Mailed check 56.95 1889.5 No 1
2 0 0 0 2 1 0 DSL Yes Yes No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes 1
3 0 0 0 45 0 0 DSL Yes No Yes Yes No No One year No Bank transfer (automatic) 42.30 1840.75 No 1
4 0 0 0 2 1 0 Fiber optic No No No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 DSL Yes No Yes Yes Yes Yes One year Yes Mailed check 84.80 1990.5 No 1
7039 0 1 1 72 1 1 Fiber optic No Yes Yes No Yes Yes One year Yes Credit card (automatic) 103.20 7362.9 No 0
7040 0 1 1 11 0 0 DSL Yes No No No No No Month-to-month Yes Electronic check 29.60 346.45 No 0
7041 1 1 0 4 1 1 Fiber optic No No No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes 1
7042 0 0 0 66 1 0 Fiber optic Yes No Yes Yes Yes Yes Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1

7043 rows × 20 columns

7. InternetService¶

In [36]:
df['InternetService'].isnull().sum()
Out[36]:
0
In [37]:
df['InternetService'].unique()
Out[37]:
array(['DSL', 'Fiber optic', 'No'], dtype=object)

Let's perform one-hot encoding for the feature 'InternetService'

In [38]:
internet_service = pd.get_dummies(df['InternetService'], prefix = 'InternetService', dtype = 'int')

Dropping the feature 'InternetService' and including encoded features from 'internet_service'

In [39]:
df = df.drop(['InternetService'], axis = 1)
In [40]:
df = pd.concat((df, internet_service), axis = 1)
In [41]:
df
Out[41]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport ... Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No
0 0 1 0 1 0 0 No Yes No No ... Month-to-month Yes Electronic check 29.85 29.85 No 0 1 0 0
1 0 0 0 34 1 0 Yes No Yes No ... One year No Mailed check 56.95 1889.5 No 1 1 0 0
2 0 0 0 2 1 0 Yes Yes No No ... Month-to-month Yes Mailed check 53.85 108.15 Yes 1 1 0 0
3 0 0 0 45 0 0 Yes No Yes Yes ... One year No Bank transfer (automatic) 42.30 1840.75 No 1 1 0 0
4 0 0 0 2 1 0 No No No No ... Month-to-month Yes Electronic check 70.70 151.65 Yes 0 0 1 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 Yes No Yes Yes ... One year Yes Mailed check 84.80 1990.5 No 1 1 0 0
7039 0 1 1 72 1 1 No Yes Yes No ... One year Yes Credit card (automatic) 103.20 7362.9 No 0 0 1 0
7040 0 1 1 11 0 0 Yes No No No ... Month-to-month Yes Electronic check 29.60 346.45 No 0 1 0 0
7041 1 1 0 4 1 1 No No No No ... Month-to-month Yes Mailed check 74.40 306.6 Yes 1 0 1 0
7042 0 0 0 66 1 0 Yes No Yes Yes ... Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1 0 1 0

7043 rows × 22 columns

8. OnlineSecurity¶

In [42]:
df['OnlineSecurity'].isnull().sum()
Out[42]:
0
In [43]:
df['OnlineSecurity'].unique()
Out[43]:
array(['No', 'Yes', 'No internet service'], dtype=object)
In [44]:
df.loc[df['OnlineSecurity'] == "No internet service", "OnlineSecurity"] = "No"

We change that 'No internet service' to 'No'. Now, we can perform Label Encoding using the object 'le'.

In [45]:
df['OnlineSecurity'].unique()
Out[45]:
array(['No', 'Yes'], dtype=object)
In [46]:
df
Out[46]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport ... Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No
0 0 1 0 1 0 0 No Yes No No ... Month-to-month Yes Electronic check 29.85 29.85 No 0 1 0 0
1 0 0 0 34 1 0 Yes No Yes No ... One year No Mailed check 56.95 1889.5 No 1 1 0 0
2 0 0 0 2 1 0 Yes Yes No No ... Month-to-month Yes Mailed check 53.85 108.15 Yes 1 1 0 0
3 0 0 0 45 0 0 Yes No Yes Yes ... One year No Bank transfer (automatic) 42.30 1840.75 No 1 1 0 0
4 0 0 0 2 1 0 No No No No ... Month-to-month Yes Electronic check 70.70 151.65 Yes 0 0 1 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 Yes No Yes Yes ... One year Yes Mailed check 84.80 1990.5 No 1 1 0 0
7039 0 1 1 72 1 1 No Yes Yes No ... One year Yes Credit card (automatic) 103.20 7362.9 No 0 0 1 0
7040 0 1 1 11 0 0 Yes No No No ... Month-to-month Yes Electronic check 29.60 346.45 No 0 1 0 0
7041 1 1 0 4 1 1 No No No No ... Month-to-month Yes Mailed check 74.40 306.6 Yes 1 0 1 0
7042 0 0 0 66 1 0 Yes No Yes Yes ... Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1 0 1 0

7043 rows × 22 columns

In [47]:
df['OnlineSecurity'] = le.fit_transform(df['OnlineSecurity'])
In [48]:
df
Out[48]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport ... Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No
0 0 1 0 1 0 0 0 Yes No No ... Month-to-month Yes Electronic check 29.85 29.85 No 0 1 0 0
1 0 0 0 34 1 0 1 No Yes No ... One year No Mailed check 56.95 1889.5 No 1 1 0 0
2 0 0 0 2 1 0 1 Yes No No ... Month-to-month Yes Mailed check 53.85 108.15 Yes 1 1 0 0
3 0 0 0 45 0 0 1 No Yes Yes ... One year No Bank transfer (automatic) 42.30 1840.75 No 1 1 0 0
4 0 0 0 2 1 0 0 No No No ... Month-to-month Yes Electronic check 70.70 151.65 Yes 0 0 1 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 1 No Yes Yes ... One year Yes Mailed check 84.80 1990.5 No 1 1 0 0
7039 0 1 1 72 1 1 0 Yes Yes No ... One year Yes Credit card (automatic) 103.20 7362.9 No 0 0 1 0
7040 0 1 1 11 0 0 1 No No No ... Month-to-month Yes Electronic check 29.60 346.45 No 0 1 0 0
7041 1 1 0 4 1 1 0 No No No ... Month-to-month Yes Mailed check 74.40 306.6 Yes 1 0 1 0
7042 0 0 0 66 1 0 1 No Yes Yes ... Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1 0 1 0

7043 rows × 22 columns

9. OnlineBackup¶

In [49]:
df['OnlineBackup'].isnull().sum()
Out[49]:
0
In [50]:
df['OnlineBackup'].unique()
Out[50]:
array(['Yes', 'No', 'No internet service'], dtype=object)
In [51]:
df.loc[df['OnlineBackup'] == "No internet service", "OnlineBackup"] = "No"
In [52]:
df['OnlineBackup'].unique()
Out[52]:
array(['Yes', 'No'], dtype=object)
In [53]:
df['OnlineBackup'] = le.fit_transform(df['OnlineBackup'])
In [54]:
df
Out[54]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport ... Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No
0 0 1 0 1 0 0 0 1 No No ... Month-to-month Yes Electronic check 29.85 29.85 No 0 1 0 0
1 0 0 0 34 1 0 1 0 Yes No ... One year No Mailed check 56.95 1889.5 No 1 1 0 0
2 0 0 0 2 1 0 1 1 No No ... Month-to-month Yes Mailed check 53.85 108.15 Yes 1 1 0 0
3 0 0 0 45 0 0 1 0 Yes Yes ... One year No Bank transfer (automatic) 42.30 1840.75 No 1 1 0 0
4 0 0 0 2 1 0 0 0 No No ... Month-to-month Yes Electronic check 70.70 151.65 Yes 0 0 1 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 1 0 Yes Yes ... One year Yes Mailed check 84.80 1990.5 No 1 1 0 0
7039 0 1 1 72 1 1 0 1 Yes No ... One year Yes Credit card (automatic) 103.20 7362.9 No 0 0 1 0
7040 0 1 1 11 0 0 1 0 No No ... Month-to-month Yes Electronic check 29.60 346.45 No 0 1 0 0
7041 1 1 0 4 1 1 0 0 No No ... Month-to-month Yes Mailed check 74.40 306.6 Yes 1 0 1 0
7042 0 0 0 66 1 0 1 0 Yes Yes ... Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1 0 1 0

7043 rows × 22 columns

10. DeviceProtection¶

In [55]:
df['DeviceProtection'].isnull().sum()
Out[55]:
0
In [56]:
df['DeviceProtection'].unique()
Out[56]:
array(['No', 'Yes', 'No internet service'], dtype=object)
In [57]:
df.loc[df['DeviceProtection'] == "No internet service", "DeviceProtection"] = "No"
In [58]:
df['DeviceProtection'].unique()
Out[58]:
array(['No', 'Yes'], dtype=object)
In [59]:
df['DeviceProtection'] = le.fit_transform(df['DeviceProtection'])
In [60]:
df
Out[60]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport ... Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No
0 0 1 0 1 0 0 0 1 0 No ... Month-to-month Yes Electronic check 29.85 29.85 No 0 1 0 0
1 0 0 0 34 1 0 1 0 1 No ... One year No Mailed check 56.95 1889.5 No 1 1 0 0
2 0 0 0 2 1 0 1 1 0 No ... Month-to-month Yes Mailed check 53.85 108.15 Yes 1 1 0 0
3 0 0 0 45 0 0 1 0 1 Yes ... One year No Bank transfer (automatic) 42.30 1840.75 No 1 1 0 0
4 0 0 0 2 1 0 0 0 0 No ... Month-to-month Yes Electronic check 70.70 151.65 Yes 0 0 1 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 1 0 1 Yes ... One year Yes Mailed check 84.80 1990.5 No 1 1 0 0
7039 0 1 1 72 1 1 0 1 1 No ... One year Yes Credit card (automatic) 103.20 7362.9 No 0 0 1 0
7040 0 1 1 11 0 0 1 0 0 No ... Month-to-month Yes Electronic check 29.60 346.45 No 0 1 0 0
7041 1 1 0 4 1 1 0 0 0 No ... Month-to-month Yes Mailed check 74.40 306.6 Yes 1 0 1 0
7042 0 0 0 66 1 0 1 0 1 Yes ... Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1 0 1 0

7043 rows × 22 columns

11. TechSupport¶

In [61]:
df['TechSupport'].isnull().sum()
Out[61]:
0
In [62]:
df['TechSupport'].unique()
Out[62]:
array(['No', 'Yes', 'No internet service'], dtype=object)
In [63]:
df.loc[df['TechSupport'] == "No internet service", "TechSupport"] = "No"
In [64]:
df['TechSupport'].unique()
Out[64]:
array(['No', 'Yes'], dtype=object)
In [65]:
df['TechSupport'] = le.fit_transform(df['TechSupport'])
In [66]:
df
Out[66]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport ... Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No
0 0 1 0 1 0 0 0 1 0 0 ... Month-to-month Yes Electronic check 29.85 29.85 No 0 1 0 0
1 0 0 0 34 1 0 1 0 1 0 ... One year No Mailed check 56.95 1889.5 No 1 1 0 0
2 0 0 0 2 1 0 1 1 0 0 ... Month-to-month Yes Mailed check 53.85 108.15 Yes 1 1 0 0
3 0 0 0 45 0 0 1 0 1 1 ... One year No Bank transfer (automatic) 42.30 1840.75 No 1 1 0 0
4 0 0 0 2 1 0 0 0 0 0 ... Month-to-month Yes Electronic check 70.70 151.65 Yes 0 0 1 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 1 0 1 1 ... One year Yes Mailed check 84.80 1990.5 No 1 1 0 0
7039 0 1 1 72 1 1 0 1 1 0 ... One year Yes Credit card (automatic) 103.20 7362.9 No 0 0 1 0
7040 0 1 1 11 0 0 1 0 0 0 ... Month-to-month Yes Electronic check 29.60 346.45 No 0 1 0 0
7041 1 1 0 4 1 1 0 0 0 0 ... Month-to-month Yes Mailed check 74.40 306.6 Yes 1 0 1 0
7042 0 0 0 66 1 0 1 0 1 1 ... Two year Yes Bank transfer (automatic) 105.65 6844.5 No 1 0 1 0

7043 rows × 22 columns

In [67]:
df.dtypes
Out[67]:
SeniorCitizen                    int64
Partner                          int32
Dependents                       int32
tenure                           int64
PhoneService                     int32
MultipleLines                    int32
OnlineSecurity                   int32
OnlineBackup                     int32
DeviceProtection                 int32
TechSupport                      int32
StreamingTV                     object
StreamingMovies                 object
Contract                        object
PaperlessBilling                object
PaymentMethod                   object
MonthlyCharges                 float64
TotalCharges                    object
Churn                           object
gender_Male                      int32
InternetService_DSL              int32
InternetService_Fiber optic      int32
InternetService_No               int32
dtype: object

12.StreamingTV¶

In [68]:
df['StreamingTV'].isnull().sum()
Out[68]:
0
In [69]:
df['StreamingTV'].unique()
Out[69]:
array(['No', 'Yes', 'No internet service'], dtype=object)
In [70]:
df.loc[df['StreamingTV'] == "No internet service", "StreamingTV"] = "No"
In [71]:
df['StreamingTV'].unique()
Out[71]:
array(['No', 'Yes'], dtype=object)
In [72]:
df['StreamingTV'] = le.fit_transform(df['StreamingTV'])
In [73]:
df['StreamingTV']
Out[73]:
0       0
1       0
2       0
3       0
4       0
       ..
7038    1
7039    1
7040    0
7041    0
7042    1
Name: StreamingTV, Length: 7043, dtype: int32

13. StreamingMovies¶

In [74]:
df['StreamingMovies'].isnull().sum()
Out[74]:
0
In [75]:
df['StreamingMovies'].unique()
Out[75]:
array(['No', 'Yes', 'No internet service'], dtype=object)
In [76]:
df.loc[df['StreamingMovies'] == "No internet service", "StreamingMovies"] = "No"
In [77]:
df['StreamingMovies'].unique()
Out[77]:
array(['No', 'Yes'], dtype=object)
In [78]:
df['StreamingMovies'] = le.fit_transform(df['StreamingMovies'])
In [79]:
df['StreamingMovies']
Out[79]:
0       0
1       0
2       0
3       0
4       0
       ..
7038    1
7039    1
7040    0
7041    0
7042    1
Name: StreamingMovies, Length: 7043, dtype: int32

14. Contract¶

In [80]:
df['Contract'].isnull().sum()
Out[80]:
0
In [81]:
df['Contract'].unique()
Out[81]:
array(['Month-to-month', 'One year', 'Two year'], dtype=object)

We've to perform one hot encoding here...

In [82]:
contract = pd.get_dummies(df['Contract'], prefix = 'Contract', dtype = 'int' )
In [83]:
contract
Out[83]:
Contract_Month-to-month Contract_One year Contract_Two year
0 1 0 0
1 0 1 0
2 1 0 0
3 0 1 0
4 1 0 0
... ... ... ...
7038 0 1 0
7039 0 1 0
7040 1 0 0
7041 1 0 0
7042 0 0 1

7043 rows × 3 columns

In [84]:
df = df.drop(['Contract'], axis = 1)
In [85]:
df = pd.concat((df, contract), axis = 1)
In [86]:
df
Out[86]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport ... MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No Contract_Month-to-month Contract_One year Contract_Two year
0 0 1 0 1 0 0 0 1 0 0 ... 29.85 29.85 No 0 1 0 0 1 0 0
1 0 0 0 34 1 0 1 0 1 0 ... 56.95 1889.5 No 1 1 0 0 0 1 0
2 0 0 0 2 1 0 1 1 0 0 ... 53.85 108.15 Yes 1 1 0 0 1 0 0
3 0 0 0 45 0 0 1 0 1 1 ... 42.30 1840.75 No 1 1 0 0 0 1 0
4 0 0 0 2 1 0 0 0 0 0 ... 70.70 151.65 Yes 0 0 1 0 1 0 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 1 0 1 1 ... 84.80 1990.5 No 1 1 0 0 0 1 0
7039 0 1 1 72 1 1 0 1 1 0 ... 103.20 7362.9 No 0 0 1 0 0 1 0
7040 0 1 1 11 0 0 1 0 0 0 ... 29.60 346.45 No 0 1 0 0 1 0 0
7041 1 1 0 4 1 1 0 0 0 0 ... 74.40 306.6 Yes 1 0 1 0 1 0 0
7042 0 0 0 66 1 0 1 0 1 1 ... 105.65 6844.5 No 1 0 1 0 0 0 1

7043 rows × 24 columns

15. PaperlessBilling¶

In [87]:
df['PaperlessBilling'].isnull().sum()
Out[87]:
0
In [88]:
df['PaperlessBilling'].unique()
Out[88]:
array(['Yes', 'No'], dtype=object)
In [89]:
df['PaperlessBilling'] = le.fit_transform(df['PaperlessBilling'])
In [90]:
df['PaperlessBilling']
Out[90]:
0       1
1       0
2       1
3       0
4       1
       ..
7038    1
7039    1
7040    1
7041    1
7042    1
Name: PaperlessBilling, Length: 7043, dtype: int32

16. PaymentMethod¶

In [91]:
df['PaymentMethod'].isnull().sum()
Out[91]:
0
In [92]:
df['PaymentMethod'].unique()
Out[92]:
array(['Electronic check', 'Mailed check', 'Bank transfer (automatic)',
       'Credit card (automatic)'], dtype=object)
In [93]:
PaymentMethod = pd.get_dummies(df['PaymentMethod'], prefix = 'PaymentMethod', dtype = 'int')
In [94]:
PaymentMethod
Out[94]:
PaymentMethod_Bank transfer (automatic) PaymentMethod_Credit card (automatic) PaymentMethod_Electronic check PaymentMethod_Mailed check
0 0 0 1 0
1 0 0 0 1
2 0 0 0 1
3 1 0 0 0
4 0 0 1 0
... ... ... ... ...
7038 0 0 0 1
7039 0 1 0 0
7040 0 0 1 0
7041 0 0 0 1
7042 1 0 0 0

7043 rows × 4 columns

In [95]:
df = df.drop(['PaymentMethod'], axis = 1)
In [96]:
df = pd.concat((df, PaymentMethod), axis = 1)
In [97]:
df
Out[97]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport ... InternetService_DSL InternetService_Fiber optic InternetService_No Contract_Month-to-month Contract_One year Contract_Two year PaymentMethod_Bank transfer (automatic) PaymentMethod_Credit card (automatic) PaymentMethod_Electronic check PaymentMethod_Mailed check
0 0 1 0 1 0 0 0 1 0 0 ... 1 0 0 1 0 0 0 0 1 0
1 0 0 0 34 1 0 1 0 1 0 ... 1 0 0 0 1 0 0 0 0 1
2 0 0 0 2 1 0 1 1 0 0 ... 1 0 0 1 0 0 0 0 0 1
3 0 0 0 45 0 0 1 0 1 1 ... 1 0 0 0 1 0 1 0 0 0
4 0 0 0 2 1 0 0 0 0 0 ... 0 1 0 1 0 0 0 0 1 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 1 0 1 1 ... 1 0 0 0 1 0 0 0 0 1
7039 0 1 1 72 1 1 0 1 1 0 ... 0 1 0 0 1 0 0 1 0 0
7040 0 1 1 11 0 0 1 0 0 0 ... 1 0 0 1 0 0 0 0 1 0
7041 1 1 0 4 1 1 0 0 0 0 ... 0 1 0 1 0 0 0 0 0 1
7042 0 0 0 66 1 0 1 0 1 1 ... 0 1 0 0 0 1 1 0 0 0

7043 rows × 27 columns

In [98]:
df.dtypes
Out[98]:
SeniorCitizen                                int64
Partner                                      int32
Dependents                                   int32
tenure                                       int64
PhoneService                                 int32
MultipleLines                                int32
OnlineSecurity                               int32
OnlineBackup                                 int32
DeviceProtection                             int32
TechSupport                                  int32
StreamingTV                                  int32
StreamingMovies                              int32
PaperlessBilling                             int32
MonthlyCharges                             float64
TotalCharges                                object
Churn                                       object
gender_Male                                  int32
InternetService_DSL                          int32
InternetService_Fiber optic                  int32
InternetService_No                           int32
Contract_Month-to-month                      int32
Contract_One year                            int32
Contract_Two year                            int32
PaymentMethod_Bank transfer (automatic)      int32
PaymentMethod_Credit card (automatic)        int32
PaymentMethod_Electronic check               int32
PaymentMethod_Mailed check                   int32
dtype: object

17. TotalCharges¶

In [99]:
df['TotalCharges'].isnull().sum()
Out[99]:
0
In [100]:
df['TotalCharges']
Out[100]:
0         29.85
1        1889.5
2        108.15
3       1840.75
4        151.65
         ...   
7038     1990.5
7039     7362.9
7040     346.45
7041      306.6
7042     6844.5
Name: TotalCharges, Length: 7043, dtype: object

From the above output, we see that all are numbers(float). But, dtype tells it is an object. So, we have to convert from object type to float.

In [101]:
 #df['TotalCharges'] = pd.to_numeric(df['TotalCharges'])

Error raises at 488 Index value.. says that failed to convert empty string as numeric... We have to handle this problem..¶

In [102]:
df['TotalCharges'][488]
Out[102]:
' '
In [103]:
df.iloc[488]
Out[103]:
SeniorCitizen                                  0
Partner                                        1
Dependents                                     1
tenure                                         0
PhoneService                                   0
MultipleLines                                  0
OnlineSecurity                                 1
OnlineBackup                                   0
DeviceProtection                               1
TechSupport                                    1
StreamingTV                                    1
StreamingMovies                                0
PaperlessBilling                               1
MonthlyCharges                             52.55
TotalCharges                                    
Churn                                         No
gender_Male                                    0
InternetService_DSL                            1
InternetService_Fiber optic                    0
InternetService_No                             0
Contract_Month-to-month                        0
Contract_One year                              0
Contract_Two year                              1
PaymentMethod_Bank transfer (automatic)        1
PaymentMethod_Credit card (automatic)          0
PaymentMethod_Electronic check                 0
PaymentMethod_Mailed check                     0
Name: 488, dtype: object

From the above information we observed that TotalCharges are not updated for this record... Let's check for all records when tenure == 0.¶

In [104]:
pd.set_option('display.max_columns', None)
In [105]:
df.where(df['tenure'] == 0).dropna()
Out[105]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies PaperlessBilling MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No Contract_Month-to-month Contract_One year Contract_Two year PaymentMethod_Bank transfer (automatic) PaymentMethod_Credit card (automatic) PaymentMethod_Electronic check PaymentMethod_Mailed check
488 0.0 1.0 1.0 0.0 0.0 0.0 1.0 0.0 1.0 1.0 1.0 0.0 1.0 52.55 No 0.0 1.0 0.0 0.0 0.0 0.0 1.0 1.0 0.0 0.0 0.0
753 0.0 0.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 20.25 No 1.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
936 0.0 1.0 1.0 0.0 1.0 0.0 1.0 1.0 1.0 0.0 1.0 1.0 0.0 80.85 No 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
1082 0.0 1.0 1.0 0.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 25.75 No 1.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
1340 0.0 1.0 1.0 0.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 56.05 No 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0 1.0 0.0 0.0
3331 0.0 1.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 19.85 No 1.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
3826 0.0 1.0 1.0 0.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 25.35 No 1.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
4380 0.0 1.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 20.00 No 0.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
5218 0.0 1.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 19.70 No 1.0 0.0 0.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 1.0
6670 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0 1.0 1.0 1.0 0.0 0.0 73.35 No 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
6754 0.0 0.0 1.0 0.0 1.0 1.0 1.0 1.0 0.0 1.0 0.0 0.0 1.0 61.90 No 1.0 1.0 0.0 0.0 0.0 0.0 1.0 1.0 0.0 0.0 0.0

Total Charges are not updated for all customers when tenure == 0. We can't update it with 0. There is a monthly charge column says that the customer have to pay some value for this month. So, we update it with that montly charges value...¶

In [106]:
to_change = df.where(df['TotalCharges'] == ' ').dropna()
In [107]:
to_change.index
Out[107]:
Index([488, 753, 936, 1082, 1340, 3331, 3826, 4380, 5218, 6670, 6754], dtype='int64')
In [108]:
for x in to_change.index:
    df.loc[df['TotalCharges'] == " ", "TotalCharges"] = df.iloc[x]['MonthlyCharges']
In [109]:
df.where(df['tenure'] == 0).dropna()
Out[109]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies PaperlessBilling MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No Contract_Month-to-month Contract_One year Contract_Two year PaymentMethod_Bank transfer (automatic) PaymentMethod_Credit card (automatic) PaymentMethod_Electronic check PaymentMethod_Mailed check
488 0.0 1.0 1.0 0.0 0.0 0.0 1.0 0.0 1.0 1.0 1.0 0.0 1.0 52.55 52.55 No 0.0 1.0 0.0 0.0 0.0 0.0 1.0 1.0 0.0 0.0 0.0
753 0.0 0.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 20.25 52.55 No 1.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
936 0.0 1.0 1.0 0.0 1.0 0.0 1.0 1.0 1.0 0.0 1.0 1.0 0.0 80.85 52.55 No 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
1082 0.0 1.0 1.0 0.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 25.75 52.55 No 1.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
1340 0.0 1.0 1.0 0.0 0.0 0.0 1.0 1.0 1.0 1.0 1.0 0.0 0.0 56.05 52.55 No 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0 1.0 0.0 0.0
3331 0.0 1.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 19.85 52.55 No 1.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
3826 0.0 1.0 1.0 0.0 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 25.35 52.55 No 1.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
4380 0.0 1.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 20.00 52.55 No 0.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
5218 0.0 1.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 19.70 52.55 No 1.0 0.0 0.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 1.0
6670 0.0 1.0 1.0 0.0 1.0 1.0 0.0 1.0 1.0 1.0 1.0 0.0 0.0 73.35 52.55 No 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
6754 0.0 0.0 1.0 0.0 1.0 1.0 1.0 1.0 0.0 1.0 0.0 0.0 1.0 61.90 52.55 No 1.0 1.0 0.0 0.0 0.0 0.0 1.0 1.0 0.0 0.0 0.0
In [110]:
df['TotalCharges'] = pd.to_numeric(df['TotalCharges'])
In [111]:
df.dtypes
Out[111]:
SeniorCitizen                                int64
Partner                                      int32
Dependents                                   int32
tenure                                       int64
PhoneService                                 int32
MultipleLines                                int32
OnlineSecurity                               int32
OnlineBackup                                 int32
DeviceProtection                             int32
TechSupport                                  int32
StreamingTV                                  int32
StreamingMovies                              int32
PaperlessBilling                             int32
MonthlyCharges                             float64
TotalCharges                               float64
Churn                                       object
gender_Male                                  int32
InternetService_DSL                          int32
InternetService_Fiber optic                  int32
InternetService_No                           int32
Contract_Month-to-month                      int32
Contract_One year                            int32
Contract_Two year                            int32
PaymentMethod_Bank transfer (automatic)      int32
PaymentMethod_Credit card (automatic)        int32
PaymentMethod_Electronic check               int32
PaymentMethod_Mailed check                   int32
dtype: object

18. Churn¶

In [112]:
df['Churn'].isnull().sum()
Out[112]:
0
In [113]:
df['Churn'].unique()
Out[113]:
array(['No', 'Yes'], dtype=object)
In [114]:
df['Churn'] = le.fit_transform(df['Churn'])
In [115]:
df
Out[115]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies PaperlessBilling MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No Contract_Month-to-month Contract_One year Contract_Two year PaymentMethod_Bank transfer (automatic) PaymentMethod_Credit card (automatic) PaymentMethod_Electronic check PaymentMethod_Mailed check
0 0 1 0 1 0 0 0 1 0 0 0 0 1 29.85 29.85 0 0 1 0 0 1 0 0 0 0 1 0
1 0 0 0 34 1 0 1 0 1 0 0 0 0 56.95 1889.50 0 1 1 0 0 0 1 0 0 0 0 1
2 0 0 0 2 1 0 1 1 0 0 0 0 1 53.85 108.15 1 1 1 0 0 1 0 0 0 0 0 1
3 0 0 0 45 0 0 1 0 1 1 0 0 0 42.30 1840.75 0 1 1 0 0 0 1 0 1 0 0 0
4 0 0 0 2 1 0 0 0 0 0 0 0 1 70.70 151.65 1 0 0 1 0 1 0 0 0 0 1 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 1 0 1 1 1 1 1 84.80 1990.50 0 1 1 0 0 0 1 0 0 0 0 1
7039 0 1 1 72 1 1 0 1 1 0 1 1 1 103.20 7362.90 0 0 0 1 0 0 1 0 0 1 0 0
7040 0 1 1 11 0 0 1 0 0 0 0 0 1 29.60 346.45 0 0 1 0 0 1 0 0 0 0 1 0
7041 1 1 0 4 1 1 0 0 0 0 0 0 1 74.40 306.60 1 1 0 1 0 1 0 0 0 0 0 1
7042 0 0 0 66 1 0 1 0 1 1 1 1 1 105.65 6844.50 0 1 0 1 0 0 0 1 1 0 0 0

7043 rows × 27 columns

In [116]:
df.dtypes
Out[116]:
SeniorCitizen                                int64
Partner                                      int32
Dependents                                   int32
tenure                                       int64
PhoneService                                 int32
MultipleLines                                int32
OnlineSecurity                               int32
OnlineBackup                                 int32
DeviceProtection                             int32
TechSupport                                  int32
StreamingTV                                  int32
StreamingMovies                              int32
PaperlessBilling                             int32
MonthlyCharges                             float64
TotalCharges                               float64
Churn                                        int32
gender_Male                                  int32
InternetService_DSL                          int32
InternetService_Fiber optic                  int32
InternetService_No                           int32
Contract_Month-to-month                      int32
Contract_One year                            int32
Contract_Two year                            int32
PaymentMethod_Bank transfer (automatic)      int32
PaymentMethod_Credit card (automatic)        int32
PaymentMethod_Electronic check               int32
PaymentMethod_Mailed check                   int32
dtype: object

Exploratory Data Aalysis¶

In [117]:
temp_df = pd.read_csv('WA_Fn-UseC_-Telco-Customer-Churn.csv')
In [118]:
temp_df.dtypes
Out[118]:
customerID           object
gender               object
SeniorCitizen         int64
Partner              object
Dependents           object
tenure                int64
PhoneService         object
MultipleLines        object
InternetService      object
OnlineSecurity       object
OnlineBackup         object
DeviceProtection     object
TechSupport          object
StreamingTV          object
StreamingMovies      object
Contract             object
PaperlessBilling     object
PaymentMethod        object
MonthlyCharges      float64
TotalCharges         object
Churn                object
dtype: object
In [119]:
import plotly.express as plx
In [120]:
fig = plx.bar(temp_df, x = 'gender', color = 'Churn')
fig.update_traces(dict(marker_line_width=0))
fig.show()

From the above bar chart, we confirm that same amount of churn occurs in both the genders¶

In [121]:
fig = plx.histogram(temp_df, x = 'SeniorCitizen', color = 'Churn')
fig.update_traces(dict(marker_line_width=0))
fig.show()

Out of 1142 SeniorCitizen 476 were churn¶

Out of 5901 Non Seniorcitizen, only 1393 were churn¶

The churn precentage are,¶

SeniorCitizen - 41.68%¶

Non-SeniorCitizen - 23.6%¶

Nearly half of the SeniorCitizen were churn.¶

In [122]:
fig = plx.histogram(temp_df, x = 'Partner', color = 'Churn')
fig.update_traces(dict(marker_line_width=0))
fig.show()

Out of 3402 partners, 669 i.e., 19.66% were churn.¶

Out of 3641 non-partners, 1200 i.e., 32.95% were churn¶

In [123]:
fig = plx.histogram(temp_df, x = 'Dependents', color = 'Churn')
fig.update_traces(dict(marker_line_width=0))
fig.show()

Out of 2110 dependents, only 326 i.e., 15.45% were churn¶

Out of 4933 non-dependents, 1543 i.e., 31.279% were churn¶

In [124]:
fig = plx.scatter(temp_df, y = 'tenure', color = 'Churn')
fig.update_traces(dict(marker_line_width=0))
fig.show()

This scatter plot is confusing a lot. So, we're going to separe the tenure value based on churn value.

In [125]:
tenure_churn_yes, tenure_churn_no = [], []
In [126]:
for i in range(len(df)):
    tenure = temp_df.iloc[i]['tenure']
    churn = temp_df.iloc[i]['Churn']
    if churn == 'Yes':
        tenure_churn_yes.append(tenure)
    else:
        tenure_churn_no.append(tenure)
In [127]:
fig = plx.scatter(y = tenure_churn_yes, title = 'Churn with tenure')
fig.update_traces(dict(marker_line_width=0))
fig.show()

Churn is high when tenure is low, i.e., Due to some reasons the customers goes away in early stages. Tenure is nothing but the period of time that he/she stays in that business/organization.¶

In [128]:
fig = plx.scatter(y = tenure_churn_no, title = 'No Churn with Tenure')
fig.update_traces(dict(marker_line_width=0))
fig.show()

From the above scatter plot, we can't observe anything because it was spreaded evenly according to the tenure.¶

In [129]:
fig = plx.histogram(temp_df, x = 'PhoneService', color = 'Churn')
fig.update_traces(dict(marker_line_width=0))
fig.show()

From the above histogram,¶

Out of 682 Citizen who not having PhoneService, 170 i.e., 24.92% were churn...¶

out of 6361 Citizen having PhoneService, 1699 i.e., 26.70 were churn...¶

In [130]:
temp_df.dtypes
Out[130]:
customerID           object
gender               object
SeniorCitizen         int64
Partner              object
Dependents           object
tenure                int64
PhoneService         object
MultipleLines        object
InternetService      object
OnlineSecurity       object
OnlineBackup         object
DeviceProtection     object
TechSupport          object
StreamingTV          object
StreamingMovies      object
Contract             object
PaperlessBilling     object
PaymentMethod        object
MonthlyCharges      float64
TotalCharges         object
Churn                object
dtype: object
In [131]:
temp_df['MultipleLines'].unique()
Out[131]:
array(['No phone service', 'No', 'Yes'], dtype=object)
In [132]:
temp_df.loc[temp_df['MultipleLines'] == 'No phone service', 'MultipleLines'] = 'No'
In [133]:
temp_df['MultipleLines'].unique()
Out[133]:
array(['No', 'Yes'], dtype=object)
In [134]:
fig = plx.histogram(temp_df, x = 'MultipleLines', color = 'Churn')
fig.update_traces(dict(marker_line_width=0))
fig.show()

From the above histogram, we see that around 25% of citizens were churn in both the scenarios i.e., citizen having mutiplelines facility and citizen not having multiple line facility.¶

In [135]:
temp_df['InternetService'].unique()
Out[135]:
array(['DSL', 'Fiber optic', 'No'], dtype=object)
In [136]:
fig = plx.histogram(temp_df, x = 'InternetService', color = 'Churn')
fig.update_traces(dict(marker_line_width=0))
fig.show()

From the above histogram, we clearly notice that nearly 42% of the FiberOptic Customers were churn. We really don't know the reason behind it.. May be poot Internet Facility¶

In [137]:
temp_df['OnlineSecurity'].unique()
Out[137]:
array(['No', 'Yes', 'No internet service'], dtype=object)
In [138]:
temp_df.loc[temp_df['OnlineSecurity'] == 'No internet service', 'OnlineSecurity'] = 'No'
In [139]:
temp_df['OnlineSecurity'].unique()
Out[139]:
array(['No', 'Yes'], dtype=object)
In [140]:
fig = plx.histogram(temp_df, x = 'OnlineSecurity', color = 'Churn')
fig.update_traces(dict(marker_line_width=0))
fig.show()

31.3% of the customer with No OnlineSecurity were churn. Churn percentage is less in the customers who having OnlineSecurity...¶

Let's check Churn on Total charges and Monthly Charges..¶

In [141]:
fig = plx.scatter(temp_df, x = 'MonthlyCharges', color = 'Churn')
fig.show()
In [142]:
fig = plx.scatter(temp_df, x = 'TotalCharges', color = 'Churn')
fig.show()

From the above two visualization, we confirm that there is no relationship between Totalcharges/Monthlycharges and Churn...¶

In [143]:
plx.imshow(df.corr(), text_auto = True, height = 1750, width = 1750)

From the above heatmap, we know that there is no highly correlated independent feature with the dependent feature.¶

Some of the features having descent correlation i.e.,¶

Contract_month_to_month - 0.4¶

Contract_Two_years - (-0.3)¶

Payment_method_Electronic_check - 0.3¶

InternetService_Fiber Optic - 0.3¶

In [144]:
temp_df['Churn'].unique()
Out[144]:
array(['No', 'Yes'], dtype=object)
In [145]:
churn_yes_df = temp_df.where((temp_df['Churn'] == 'Yes')).dropna()
churn_yes_df
Out[145]:
customerID gender SeniorCitizen Partner Dependents tenure PhoneService MultipleLines InternetService OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies Contract PaperlessBilling PaymentMethod MonthlyCharges TotalCharges Churn
2 3668-QPYBK Male 0.0 No No 2.0 Yes No DSL Yes Yes No No No No Month-to-month Yes Mailed check 53.85 108.15 Yes
4 9237-HQITU Female 0.0 No No 2.0 Yes No Fiber optic No No No No No No Month-to-month Yes Electronic check 70.70 151.65 Yes
5 9305-CDSKC Female 0.0 No No 8.0 Yes Yes Fiber optic No No Yes No Yes Yes Month-to-month Yes Electronic check 99.65 820.5 Yes
8 7892-POOKP Female 0.0 Yes No 28.0 Yes Yes Fiber optic No No Yes Yes Yes Yes Month-to-month Yes Electronic check 104.80 3046.05 Yes
13 0280-XJGEX Male 0.0 No No 49.0 Yes Yes Fiber optic No Yes Yes No Yes Yes Month-to-month Yes Bank transfer (automatic) 103.70 5036.3 Yes
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7021 1699-HPSBG Male 0.0 No No 12.0 Yes No DSL No No No Yes Yes No One year Yes Electronic check 59.80 727.8 Yes
7026 8775-CEBBJ Female 0.0 No No 9.0 Yes No DSL No No No No No No Month-to-month Yes Bank transfer (automatic) 44.20 403.35 Yes
7032 6894-LFHLY Male 1.0 No No 1.0 Yes Yes Fiber optic No No No No No No Month-to-month Yes Electronic check 75.75 75.75 Yes
7034 0639-TSIQW Female 0.0 No No 67.0 Yes Yes Fiber optic Yes Yes Yes No Yes No Month-to-month Yes Credit card (automatic) 102.95 6886.25 Yes
7041 8361-LTMKD Male 1.0 Yes No 4.0 Yes Yes Fiber optic No No No No No No Month-to-month Yes Mailed check 74.40 306.6 Yes

1869 rows × 21 columns

Thereare 1869 churn customers...

In [146]:
churn_yes_df['Dependents'].value_counts()
Out[146]:
Dependents
No     1543
Yes     326
Name: count, dtype: int64
In [147]:
fig = plx.histogram(churn_yes_df, x = 'Dependents', title = "Churn on Dependents")
fig.update_traces(dict(marker_line_width=0))
fig.show()

From the above bar chart, most of churn are non dependents...¶

In [148]:
churn_yes_df['tenure'].value_counts()
Out[148]:
tenure
1.0     380
2.0     123
3.0      94
4.0      83
5.0      64
       ... 
60.0      6
72.0      6
62.0      5
64.0      4
63.0      4
Name: count, Length: 72, dtype: int64
In [149]:
plx.histogram(x = churn_yes_df['tenure'])

This Histogram says that the customers were moving out in early stages...¶

In [150]:
churn_yes_df['PhoneService'].value_counts()
Out[150]:
PhoneService
Yes    1699
No      170
Name: count, dtype: int64
In [151]:
fig = plx.histogram(churn_yes_df, x = 'PhoneService', title = "Churn on PhoneService")
fig.update_traces(dict(marker_line_width=0))
fig.show()

Customers are moving out eventhough having PhoneService. Something Fishy!!!.........¶

In [152]:
churn_yes_df['InternetService'].value_counts()
Out[152]:
InternetService
Fiber optic    1297
DSL             459
No              113
Name: count, dtype: int64
In [153]:
fig = plx.histogram(churn_yes_df, x = 'InternetService', title = "Churn on InternetService")
fig.update_traces(dict(marker_line_width=0))
fig.show()

Customers with Fiber Optic connection were moving out.. This may be of poor service in Fiber Optics InternetService¶

In [154]:
churn_yes_df['Contract'].value_counts()
Out[154]:
Contract
Month-to-month    1655
One year           166
Two year            48
Name: count, dtype: int64
In [155]:
fig = plx.histogram(churn_yes_df, x = 'Contract', title = "Churn on Contract")
fig.update_traces(dict(marker_line_width=0))
fig.show()

Out of 1869 churns, 1655 were Month-to-Month contract customers...¶

In [156]:
churn_yes_df['PaymentMethod'].value_counts()
Out[156]:
PaymentMethod
Electronic check             1071
Mailed check                  308
Bank transfer (automatic)     258
Credit card (automatic)       232
Name: count, dtype: int64
In [157]:
fig = plx.histogram(churn_yes_df, x = 'PaymentMethod', title = "PaymentMethod")
fig.update_traces(dict(marker_line_width=0))
fig.show()

Out of 1869 churns, 1071 were using Payment method as Electronic Check...¶

Feature Scaling¶

In [158]:
df
Out[158]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies PaperlessBilling MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No Contract_Month-to-month Contract_One year Contract_Two year PaymentMethod_Bank transfer (automatic) PaymentMethod_Credit card (automatic) PaymentMethod_Electronic check PaymentMethod_Mailed check
0 0 1 0 1 0 0 0 1 0 0 0 0 1 29.85 29.85 0 0 1 0 0 1 0 0 0 0 1 0
1 0 0 0 34 1 0 1 0 1 0 0 0 0 56.95 1889.50 0 1 1 0 0 0 1 0 0 0 0 1
2 0 0 0 2 1 0 1 1 0 0 0 0 1 53.85 108.15 1 1 1 0 0 1 0 0 0 0 0 1
3 0 0 0 45 0 0 1 0 1 1 0 0 0 42.30 1840.75 0 1 1 0 0 0 1 0 1 0 0 0
4 0 0 0 2 1 0 0 0 0 0 0 0 1 70.70 151.65 1 0 0 1 0 1 0 0 0 0 1 0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0 1 1 24 1 1 1 0 1 1 1 1 1 84.80 1990.50 0 1 1 0 0 0 1 0 0 0 0 1
7039 0 1 1 72 1 1 0 1 1 0 1 1 1 103.20 7362.90 0 0 0 1 0 0 1 0 0 1 0 0
7040 0 1 1 11 0 0 1 0 0 0 0 0 1 29.60 346.45 0 0 1 0 0 1 0 0 0 0 1 0
7041 1 1 0 4 1 1 0 0 0 0 0 0 1 74.40 306.60 1 1 0 1 0 1 0 0 0 0 0 1
7042 0 0 0 66 1 0 1 0 1 1 1 1 1 105.65 6844.50 0 1 0 1 0 0 0 1 1 0 0 0

7043 rows × 27 columns

Let's perform a scaling in the DataFrame to bring all the features in the same scale...¶

In [159]:
from sklearn.preprocessing import MinMaxScaler
In [160]:
mms = MinMaxScaler()
In [161]:
scaled_df = mms.fit_transform(df)
In [162]:
scaled_df = pd.DataFrame(scaled_df, columns = df.columns)
In [163]:
scaled_df
Out[163]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies PaperlessBilling MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No Contract_Month-to-month Contract_One year Contract_Two year PaymentMethod_Bank transfer (automatic) PaymentMethod_Credit card (automatic) PaymentMethod_Electronic check PaymentMethod_Mailed check
0 0.0 1.0 0.0 0.013889 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.115423 0.001275 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0
1 0.0 0.0 0.0 0.472222 1.0 0.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.385075 0.215867 0.0 1.0 1.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 1.0
2 0.0 0.0 0.0 0.027778 1.0 0.0 1.0 1.0 0.0 0.0 0.0 0.0 1.0 0.354229 0.010310 1.0 1.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 1.0
3 0.0 0.0 0.0 0.625000 0.0 0.0 1.0 0.0 1.0 1.0 0.0 0.0 0.0 0.239303 0.210241 0.0 1.0 1.0 0.0 0.0 0.0 1.0 0.0 1.0 0.0 0.0 0.0
4 0.0 0.0 0.0 0.027778 1.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.521891 0.015330 1.0 0.0 0.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
7038 0.0 1.0 1.0 0.333333 1.0 1.0 1.0 0.0 1.0 1.0 1.0 1.0 1.0 0.662189 0.227521 0.0 1.0 1.0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 1.0
7039 0.0 1.0 1.0 1.000000 1.0 1.0 0.0 1.0 1.0 0.0 1.0 1.0 1.0 0.845274 0.847461 0.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0
7040 0.0 1.0 1.0 0.152778 0.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 1.0 0.112935 0.037809 0.0 0.0 1.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0 1.0 0.0
7041 1.0 1.0 0.0 0.055556 1.0 1.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.558706 0.033210 1.0 1.0 0.0 1.0 0.0 1.0 0.0 0.0 0.0 0.0 0.0 1.0
7042 0.0 0.0 0.0 0.916667 1.0 0.0 1.0 0.0 1.0 1.0 1.0 1.0 1.0 0.869652 0.787641 0.0 1.0 0.0 1.0 0.0 0.0 0.0 1.0 1.0 0.0 0.0 0.0

7043 rows × 27 columns

Now, everything sacled in the range of 0 to 1.

Creating a Model¶

Model 1¶

In [164]:
from keras.layers import Dense,Dropout
from keras import Sequential
In [165]:
ANN_model = Sequential()
In [166]:
# Adding Input Layer to ANN
ANN_model.add(Dense(units = 27, activation = 'relu'))
In [167]:
# Adding 1st Hidden Layer to the ANN
ANN_model.add(Dense(units = 15, activation = 'relu'))
ANN_model.add(Dropout(0.4))
In [168]:
# Adding 2nd Hidden Layer to the ANN
ANN_model.add(Dense(units = 7, activation = 'relu'))
ANN_model.add(Dropout(0.3))
In [169]:
# Adding Output Layer to the ANN
ANN_model.add(Dense(units = 1, activation = 'sigmoid'))
In [170]:
ANN_model.compile(optimizer = 'adam',
                  loss = 'binary_crossentropy',
                  metrics = ['accuracy'])
In [171]:
x_train, x_test, y_train, y_test = train_test_split( df.drop(['Churn'], axis = 1), df['Churn'], test_size = 0.2, random_state = 35)
In [172]:
import tensorflow as tf
early_stopping = tf.keras.callbacks.EarlyStopping(
    monitor="accuracy",
    min_delta=0.0001,
    patience=10,
    verbose=1,
    mode="auto",
    baseline=None,
    restore_best_weights=False
)
In [173]:
model_history = ANN_model.fit(x_train, y_train, batch_size = 10, epochs = 40, validation_split = 0.33, callbacks = early_stopping )

Epoch 1/40
378/378 [==============================] - 3s 5ms/step - loss: 30.6997 - accuracy: 0.6017 - val_loss: 0.6224 - val_accuracy: 0.7511
Epoch 2/40
378/378 [==============================] - 2s 4ms/step - loss: 0.6164 - accuracy: 0.7509 - val_loss: 0.5665 - val_accuracy: 0.7597
Epoch 3/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5836 - accuracy: 0.7459 - val_loss: 0.5603 - val_accuracy: 0.7527
Epoch 4/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5691 - accuracy: 0.7480 - val_loss: 0.5598 - val_accuracy: 0.7462
Epoch 5/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5606 - accuracy: 0.7509 - val_loss: 0.5419 - val_accuracy: 0.7608
Epoch 6/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5655 - accuracy: 0.7459 - val_loss: 0.5518 - val_accuracy: 0.7527
Epoch 7/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5602 - accuracy: 0.7533 - val_loss: 0.5474 - val_accuracy: 0.7532
Epoch 8/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5630 - accuracy: 0.7509 - val_loss: 0.5282 - val_accuracy: 0.7672
Epoch 9/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5728 - accuracy: 0.7456 - val_loss: 0.5604 - val_accuracy: 0.7468
Epoch 10/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5674 - accuracy: 0.7435 - val_loss: 0.5613 - val_accuracy: 0.7452
Epoch 11/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5646 - accuracy: 0.7443 - val_loss: 0.5588 - val_accuracy: 0.7478
Epoch 12/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5643 - accuracy: 0.7440 - val_loss: 0.5456 - val_accuracy: 0.7543
Epoch 13/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5735 - accuracy: 0.7403 - val_loss: 0.5574 - val_accuracy: 0.7457
Epoch 14/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5646 - accuracy: 0.7472 - val_loss: 0.5627 - val_accuracy: 0.7478
Epoch 15/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5641 - accuracy: 0.7451 - val_loss: 0.5572 - val_accuracy: 0.7473
Epoch 16/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5660 - accuracy: 0.7462 - val_loss: 0.5556 - val_accuracy: 0.7478
Epoch 17/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5663 - accuracy: 0.7417 - val_loss: 0.5583 - val_accuracy: 0.7489
Epoch 17: early stopping

There is no improvement in the accuracy for longer time. And our final train accuracy of the is 74.17%.¶

In [174]:
model_history.history['val_accuracy']
Out[174]:
[0.751075267791748,
 0.7596774101257324,
 0.7526881694793701,
 0.7462365627288818,
 0.7607526779174805,
 0.7526881694793701,
 0.7532258033752441,
 0.7672042846679688,
 0.7467741966247559,
 0.7451612949371338,
 0.7478494644165039,
 0.7543010711669922,
 0.7456989288330078,
 0.7478494644165039,
 0.7473118305206299,
 0.7478494644165039,
 0.748924732208252]
In [175]:
model_history.history['accuracy']
Out[175]:
[0.6017488241195679,
 0.7509273886680603,
 0.7458929419517517,
 0.7480127215385437,
 0.7509273886680603,
 0.7458929419517517,
 0.7533121109008789,
 0.7509273886680603,
 0.7456279993057251,
 0.7435082197189331,
 0.7443031072616577,
 0.7440381646156311,
 0.7403285503387451,
 0.7472178339958191,
 0.7450980544090271,
 0.7461579442024231,
 0.7416534423828125]
In [176]:
model_history.history.keys()
Out[176]:
dict_keys(['loss', 'accuracy', 'val_loss', 'val_accuracy'])

In the variable model_history, we recorded all information of each iteration(epochs)..¶

In [177]:
import plotly.graph_objects as go
In [240]:
fig_1 = go.Figure()
In [241]:
fig_1.add_trace(go.Scatter(x =np.arange(0,len(model_history.history['accuracy'])),
                         y = model_history.history['val_accuracy'],
                         mode='lines+markers',
                         name='val_accuracy'))
fig_1.add_trace(go.Scatter(x =np.arange(0,len(model_history.history['accuracy'])),
                         y = model_history.history['accuracy'],
                         mode='lines+markers',
                         name='Accuracy'))
fig_1.update_layout(title = 'ACCURACY vs VALIDATION_ACCURACY')

fig_1.update_xaxes(title_text="Epochs")
fig_1.update_yaxes(title_text="Accuracy")

fig_1.show()

From the above line chart, we see that there is no difference in the Accuracy and Validation_Accuracy after few epochs.¶

In [180]:
model_history.history['loss']
Out[180]:
[30.699697494506836,
 0.6163778901100159,
 0.5835999250411987,
 0.5690638422966003,
 0.5605646371841431,
 0.5655323266983032,
 0.5602178573608398,
 0.5630198121070862,
 0.5727599859237671,
 0.56740403175354,
 0.5646049380302429,
 0.5643322467803955,
 0.5735149383544922,
 0.5646213889122009,
 0.5640552043914795,
 0.5659602880477905,
 0.5663062930107117]
In [181]:
model_history.history['val_loss']
Out[181]:
[0.6224174499511719,
 0.566506028175354,
 0.5603281259536743,
 0.5597817301750183,
 0.541858434677124,
 0.5517999529838562,
 0.5473867654800415,
 0.5281704068183899,
 0.5603865385055542,
 0.5613142251968384,
 0.558760404586792,
 0.5455719232559204,
 0.5573979616165161,
 0.5626907348632812,
 0.5572342276573181,
 0.555553674697876,
 0.5582592487335205]
In [238]:
fig_2 = go.Figure()
In [239]:
fig_2.add_trace(go.Scatter(x =np.arange(0,len(model_history.history['loss'])),
                         y = model_history.history['loss'],
                         mode='lines+markers',
                         name='loss'))
fig_2.add_trace(go.Scatter(x =np.arange(0,len(model_history.history['loss'])),
                         y = model_history.history['val_loss'],
                         mode='lines+markers',
                         name='val_loss'))
fig_2.update_layout(title = 'LOSS vs VALIDATION_LOSS')

fig_2.update_xaxes(title_text="Epochs")
fig_2.update_yaxes(title_text="Loss")

fig_2.show()

From the above line chart, we see that there is no difference in the Validation_Loss and Loss after few epochs.¶

In [184]:
ANN_model.evaluate(x_test, y_test)

45/45 [==============================] - 0s 3ms/step - loss: 0.5267 - accuracy: 0.7786
Out[184]:
[0.5266959071159363, 0.7785663604736328]
In [185]:
from sklearn.metrics import confusion_matrix, classification_report
In [186]:
predict = ANN_model.predict(x_test)

45/45 [==============================] - 0s 1ms/step
In [187]:
predict
Out[187]:
array([[0.25292522],
       [0.25292522],
       [0.25292522],
       ...,
       [0.38886058],
       [0.25292522],
       [0.25292522]], dtype=float32)
In [188]:
predict_new = []
for x in predict:
    if x >= 0.5:
        predict_new.append(1)
    else:
        predict_new.append(0)
In [189]:
predict_new[-10 : ]
Out[189]:
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
In [190]:
plx.imshow(confusion_matrix( y_test, predict_new), text_auto = True)
In [191]:
print(classification_report(y_test, predict_new))

              precision    recall  f1-score   support

           0       0.77      1.00      0.87      1062
           1       0.91      0.11      0.20       347

    accuracy                           0.78      1409
   macro avg       0.84      0.55      0.54      1409
weighted avg       0.81      0.78      0.71      1409

Final test Accuracy for our 1st model is 78%¶

Model 2¶

In [192]:
plx.imshow(df.corr(), height = 1700, width = 1700, text_auto = True)

Here, we're having too many feature to predict. Some of the feature's correlation are nearly zero. It may leads to overfitting. So, we have to remove some less important features in the dataframe.¶

In [193]:
corr = df.corr()
In [194]:
df.corr()
Out[194]:
SeniorCitizen Partner Dependents tenure PhoneService MultipleLines OnlineSecurity OnlineBackup DeviceProtection TechSupport StreamingTV StreamingMovies PaperlessBilling MonthlyCharges TotalCharges Churn gender_Male InternetService_DSL InternetService_Fiber optic InternetService_No Contract_Month-to-month Contract_One year Contract_Two year PaymentMethod_Bank transfer (automatic) PaymentMethod_Credit card (automatic) PaymentMethod_Electronic check PaymentMethod_Mailed check
SeniorCitizen 1.000000 0.016479 -0.211185 0.016567 0.008576 0.142948 -0.038653 0.066572 0.059428 -0.060625 0.105378 0.120176 0.156530 0.220173 0.102994 0.150889 -0.001874 -0.108322 0.255338 -0.182742 0.138360 -0.046262 -0.117000 -0.016159 -0.024135 0.171718 -0.153477
Partner 0.016479 1.000000 0.452676 0.379697 0.017706 0.142057 0.143106 0.141498 0.153786 0.119999 0.124666 0.117412 -0.014877 0.096848 0.317540 -0.150448 -0.001808 -0.000851 0.000304 0.000615 -0.280865 0.082783 0.248091 0.110706 0.082029 -0.083852 -0.095125
Dependents -0.211185 0.452676 1.000000 0.159712 -0.001762 -0.024526 0.080972 0.023671 0.013963 0.063268 -0.016558 -0.039741 -0.111377 -0.113890 0.062136 -0.164221 0.010517 0.052010 -0.165818 0.139812 -0.231720 0.068368 0.204613 0.052021 0.060267 -0.150642 0.059071
tenure 0.016567 0.379697 0.159712 1.000000 0.008448 0.331941 0.327203 0.360277 0.360653 0.324221 0.279756 0.286111 0.006152 0.247900 0.826160 -0.352229 0.005106 0.013274 0.019720 -0.039062 -0.645561 0.202570 0.558533 0.243510 0.233006 -0.208363 -0.233852
PhoneService 0.008576 0.017706 -0.001762 0.008448 1.000000 0.279690 -0.092893 -0.052312 -0.071227 -0.096340 -0.022574 -0.032959 0.016505 0.247398 0.113207 0.011942 -0.006488 -0.452425 0.289999 0.172209 -0.000742 -0.002791 0.003519 0.007556 -0.007721 0.003062 -0.003319
MultipleLines 0.142948 0.142057 -0.024526 0.331941 0.279690 1.000000 0.098108 0.202237 0.201137 0.100571 0.257152 0.258751 0.163530 0.490434 0.468516 0.040102 -0.008414 -0.199920 0.366083 -0.210564 -0.088203 -0.003794 0.106253 0.075527 0.060048 0.083618 -0.227206
OnlineSecurity -0.038653 0.143106 0.080972 0.327203 -0.092893 0.098108 1.000000 0.283832 0.275438 0.354931 0.176207 0.187398 -0.003636 0.296594 0.411672 -0.171226 -0.017021 0.321269 -0.030696 -0.333403 -0.246679 0.100162 0.191773 0.095158 0.115721 -0.112338 -0.080798
OnlineBackup 0.066572 0.141498 0.023671 0.360277 -0.052312 0.202237 0.283832 1.000000 0.303546 0.294233 0.282106 0.274501 0.126735 0.441780 0.509246 -0.082255 -0.013773 0.157884 0.165651 -0.381593 -0.164172 0.083722 0.111400 0.087004 0.090785 -0.000408 -0.174164
DeviceProtection 0.059428 0.153786 0.013963 0.360653 -0.071227 0.201137 0.275438 0.303546 1.000000 0.333313 0.390874 0.402111 0.103797 0.482692 0.522003 -0.066160 -0.002105 0.146291 0.176049 -0.380754 -0.225662 0.102495 0.165096 0.083115 0.111554 -0.003351 -0.187373
TechSupport -0.060625 0.119999 0.063268 0.324221 -0.096340 0.100571 0.354931 0.294233 0.333313 1.000000 0.278070 0.279358 0.037880 0.338304 0.431904 -0.164674 -0.009212 0.313118 -0.020492 -0.336298 -0.285241 0.095775 0.240824 0.101252 0.117272 -0.114839 -0.085509
StreamingTV 0.105378 0.124666 -0.016558 0.279756 -0.022574 0.257152 0.176207 0.282106 0.390874 0.278070 1.000000 0.533094 0.223841 0.629603 0.514990 0.063228 -0.008393 0.016274 0.329349 -0.415552 -0.112282 0.061612 0.072049 0.046252 0.040433 0.144626 -0.247742
StreamingMovies 0.120176 0.117412 -0.039741 0.286111 -0.032959 0.258751 0.187398 0.274501 0.402111 0.279358 0.533094 1.000000 0.211716 0.627429 0.520118 0.061382 -0.010487 0.025698 0.322923 -0.418675 -0.116633 0.064926 0.073960 0.048652 0.048575 0.137966 -0.250595
PaperlessBilling 0.156530 -0.014877 -0.111377 0.006152 0.016505 0.163530 -0.003636 0.126735 0.103797 0.037880 0.223841 0.211716 1.000000 0.352150 0.158557 0.191825 -0.011754 -0.063121 0.326853 -0.321013 0.169096 -0.051391 -0.147889 -0.016332 -0.013589 0.208865 -0.205398
MonthlyCharges 0.220173 0.096848 -0.113890 0.247900 0.247398 0.490434 0.296594 0.441780 0.482692 0.338304 0.629603 0.627429 0.352150 1.000000 0.651169 0.193356 -0.014569 -0.160189 0.787066 -0.763557 0.060165 0.004904 -0.074681 0.042812 0.030550 0.271625 -0.377437
TotalCharges 0.102994 0.317540 0.062136 0.826160 0.113207 0.468516 0.411672 0.509246 0.522003 0.431904 0.514990 0.520118 0.158557 0.651169 1.000000 -0.198353 -0.000077 -0.052462 0.361636 -0.375207 -0.444311 0.170810 0.354550 0.185990 0.182910 -0.059274 -0.295726
Churn 0.150889 -0.150448 -0.164221 -0.352229 0.011942 0.040102 -0.171226 -0.082255 -0.066160 -0.164674 0.063228 0.061382 0.191825 0.193356 -0.198353 1.000000 -0.008612 -0.124214 0.308020 -0.227890 0.405103 -0.177820 -0.302253 -0.117937 -0.134302 0.301919 -0.091683
gender_Male -0.001874 -0.001808 0.010517 0.005106 -0.006488 -0.008414 -0.017021 -0.013773 -0.002105 -0.009212 -0.008393 -0.010487 -0.011754 -0.014569 -0.000077 -0.008612 1.000000 0.006568 -0.011286 0.006026 -0.003386 0.008026 -0.003695 -0.016024 0.001215 0.000752 0.013744
InternetService_DSL -0.108322 -0.000851 0.052010 0.013274 -0.452425 -0.199920 0.321269 0.157884 0.146291 0.313118 0.016274 0.025698 -0.063121 -0.160189 -0.052462 -0.124214 0.006568 1.000000 -0.640987 -0.380635 -0.065509 0.046795 0.031714 0.025476 0.051438 -0.104418 0.041899
InternetService_Fiber optic 0.255338 0.000304 -0.165818 0.019720 0.289999 0.366083 -0.030696 0.165651 0.176049 -0.020492 0.329349 0.322923 0.326853 0.787066 0.361636 0.308020 -0.011286 -0.640987 1.000000 -0.465793 0.244164 -0.076324 -0.211526 -0.022624 -0.050077 0.336410 -0.306834
InternetService_No -0.182742 0.000615 0.139812 -0.039062 0.172209 -0.210564 -0.333403 -0.381593 -0.380754 -0.336298 -0.415552 -0.418675 -0.321013 -0.763557 -0.375207 -0.227890 0.006026 -0.380635 -0.465793 1.000000 -0.218639 0.038004 0.218278 -0.002113 0.001030 -0.284917 0.321361
Contract_Month-to-month 0.138360 -0.280865 -0.231720 -0.645561 -0.000742 -0.088203 -0.246679 -0.164172 -0.225662 -0.285241 -0.112282 -0.116633 0.169096 0.060165 -0.444311 0.405103 -0.003386 -0.065509 0.244164 -0.218639 1.000000 -0.568744 -0.622633 -0.179707 -0.204145 0.331661 0.004138
Contract_One year -0.046262 0.082783 0.068368 0.202570 -0.002791 -0.003794 0.100162 0.083722 0.102495 0.095775 0.061612 0.064926 -0.051391 0.004904 0.170810 -0.177820 0.008026 0.046795 -0.076324 0.038004 -0.568744 1.000000 -0.289510 0.057451 0.067589 -0.109130 -0.000116
Contract_Two year -0.117000 0.248091 0.204613 0.558533 0.003519 0.106253 0.191773 0.111400 0.165096 0.240824 0.072049 0.073960 -0.147889 -0.074681 0.354550 -0.302253 -0.003695 0.031714 -0.211526 0.218278 -0.622633 -0.289510 1.000000 0.154471 0.173265 -0.282138 -0.004705
PaymentMethod_Bank transfer (automatic) -0.016159 0.110706 0.052021 0.243510 0.007556 0.075527 0.095158 0.087004 0.083115 0.101252 0.046252 0.048652 -0.016332 0.042812 0.185990 -0.117937 -0.016024 0.025476 -0.022624 -0.002113 -0.179707 0.057451 0.154471 1.000000 -0.278215 -0.376762 -0.288685
PaymentMethod_Credit card (automatic) -0.024135 0.082029 0.060267 0.233006 -0.007721 0.060048 0.115721 0.090785 0.111554 0.117272 0.040433 0.048575 -0.013589 0.030550 0.182910 -0.134302 0.001215 0.051438 -0.050077 0.001030 -0.204145 0.067589 0.173265 -0.278215 1.000000 -0.373322 -0.286049
PaymentMethod_Electronic check 0.171718 -0.083852 -0.150642 -0.208363 0.003062 0.083618 -0.112338 -0.000408 -0.003351 -0.114839 0.144626 0.137966 0.208865 0.271625 -0.059274 0.301919 0.000752 -0.104418 0.336410 -0.284917 0.331661 -0.109130 -0.282138 -0.376762 -0.373322 1.000000 -0.387372
PaymentMethod_Mailed check -0.153477 -0.095125 0.059071 -0.233852 -0.003319 -0.227206 -0.080798 -0.174164 -0.187373 -0.085509 -0.247742 -0.250595 -0.205398 -0.377437 -0.295726 -0.091683 0.013744 0.041899 -0.306834 0.321361 0.004138 -0.000116 -0.004705 -0.288685 -0.286049 -0.387372 1.000000
In [195]:
index = corr['Churn'].index
values = corr['Churn'].values
values = [abs(x) for x in values]
In [196]:
sort_df = pd.DataFrame()
sort_df['index'] , sort_df['values'] = list(index), list(values)
In [197]:
sort_df
Out[197]:
index values
0 SeniorCitizen 0.150889
1 Partner 0.150448
2 Dependents 0.164221
3 tenure 0.352229
4 PhoneService 0.011942
5 MultipleLines 0.040102
6 OnlineSecurity 0.171226
7 OnlineBackup 0.082255
8 DeviceProtection 0.066160
9 TechSupport 0.164674
10 StreamingTV 0.063228
11 StreamingMovies 0.061382
12 PaperlessBilling 0.191825
13 MonthlyCharges 0.193356
14 TotalCharges 0.198353
15 Churn 1.000000
16 gender_Male 0.008612
17 InternetService_DSL 0.124214
18 InternetService_Fiber optic 0.308020
19 InternetService_No 0.227890
20 Contract_Month-to-month 0.405103
21 Contract_One year 0.177820
22 Contract_Two year 0.302253
23 PaymentMethod_Bank transfer (automatic) 0.117937
24 PaymentMethod_Credit card (automatic) 0.134302
25 PaymentMethod_Electronic check 0.301919
26 PaymentMethod_Mailed check 0.091683
In [198]:
sort_df = sort_df.sort_values(by=['values'])
In [199]:
# sort_df is sorted by the column 'values'
sort_df.reset_index(inplace = True)

In the main dataframe(df), we're going to remove first 17 columns from sort_df.¶

In [200]:
to_remove_features = list(sort_df['index'][0 : 17])
In [228]:
to_remove_features
Out[228]:
['gender_Male',
 'PhoneService',
 'MultipleLines',
 'StreamingMovies',
 'StreamingTV',
 'DeviceProtection',
 'OnlineBackup',
 'PaymentMethod_Mailed check',
 'PaymentMethod_Bank transfer (automatic)',
 'InternetService_DSL',
 'PaymentMethod_Credit card (automatic)',
 'Partner',
 'SeniorCitizen',
 'Dependents',
 'TechSupport',
 'OnlineSecurity',
 'Contract_One year']

These are the 17 least important features of the dataset

In [202]:
df = df.drop(to_remove_features, axis = 1)
In [203]:
df
Out[203]:
tenure PaperlessBilling MonthlyCharges TotalCharges Churn InternetService_Fiber optic InternetService_No Contract_Month-to-month Contract_Two year PaymentMethod_Electronic check
0 1 1 29.85 29.85 0 0 0 1 0 1
1 34 0 56.95 1889.50 0 0 0 0 0 0
2 2 1 53.85 108.15 1 0 0 1 0 0
3 45 0 42.30 1840.75 0 0 0 0 0 0
4 2 1 70.70 151.65 1 1 0 1 0 1
... ... ... ... ... ... ... ... ... ... ...
7038 24 1 84.80 1990.50 0 0 0 0 0 0
7039 72 1 103.20 7362.90 0 1 0 0 0 0
7040 11 1 29.60 346.45 0 0 0 1 0 1
7041 4 1 74.40 306.60 1 1 0 1 0 0
7042 66 1 105.65 6844.50 0 1 0 0 1 0

7043 rows × 10 columns

In [204]:
x_train = x_train.drop(to_remove_features, axis = 1)
x_test = x_test.drop(to_remove_features, axis = 1)

Now, the 17 least important columns were removed.¶

In [205]:
ANN_model_2 = Sequential()
In [206]:
# Adding Input Layer to ANN
ANN_model_2.add(Dense(units = 9, activation = 'relu'))
In [207]:
# Adding 1st Hidden Layer to the ANN
ANN_model_2.add(Dense(units = 7, activation = 'relu'))
#ANN_model_2.add(Dropout(0.3))
In [208]:
# Adding 2nd Hidden Layer to the ANN
ANN_model_2.add(Dense(units = 3, activation = 'relu'))
#ANN_model_2.add(Dropout(0.3))
In [209]:
# Adding Output Layer to the ANN
ANN_model_2.add(Dense(units = 1, activation = 'sigmoid'))
In [210]:
# Compiling the ANN model with required parameters
ANN_model_2.compile(optimizer = 'adam',
                  loss = 'binary_crossentropy',
                  metrics = ['accuracy'])
In [211]:
# Early Stopping is provided to avoid running too many epochs of no improvement.
early_stopping = tf.keras.callbacks.EarlyStopping(
    monitor="accuracy",
    min_delta=0.0001,
    patience=10,
    verbose=1,
    mode="auto",
    baseline=None,
    restore_best_weights=True
)
In [212]:
model_history_2 = ANN_model_2.fit(x_train, y_train, batch_size = 10, epochs = 40, validation_split = 0.33, callbacks = early_stopping )

Epoch 1/40
378/378 [==============================] - 3s 6ms/step - loss: 0.7003 - accuracy: 0.7395 - val_loss: 0.6337 - val_accuracy: 0.7511
Epoch 2/40
378/378 [==============================] - 2s 4ms/step - loss: 0.6113 - accuracy: 0.7576 - val_loss: 0.5985 - val_accuracy: 0.7505
Epoch 3/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5821 - accuracy: 0.7546 - val_loss: 0.5769 - val_accuracy: 0.7505
Epoch 4/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5731 - accuracy: 0.7589 - val_loss: 0.5674 - val_accuracy: 0.7500
Epoch 5/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5555 - accuracy: 0.7578 - val_loss: 0.5559 - val_accuracy: 0.7516
Epoch 6/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5480 - accuracy: 0.7583 - val_loss: 0.5513 - val_accuracy: 0.7575
Epoch 7/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5446 - accuracy: 0.7570 - val_loss: 0.5348 - val_accuracy: 0.7624
Epoch 8/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5399 - accuracy: 0.7626 - val_loss: 0.5498 - val_accuracy: 0.7581
Epoch 9/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5275 - accuracy: 0.7692 - val_loss: 0.5290 - val_accuracy: 0.7710
Epoch 10/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5216 - accuracy: 0.7700 - val_loss: 0.5331 - val_accuracy: 0.7677
Epoch 11/40
378/378 [==============================] - 2s 7ms/step - loss: 0.5176 - accuracy: 0.7692 - val_loss: 0.5346 - val_accuracy: 0.7656
Epoch 12/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5181 - accuracy: 0.7684 - val_loss: 0.5259 - val_accuracy: 0.7677
Epoch 13/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5212 - accuracy: 0.7684 - val_loss: 0.5511 - val_accuracy: 0.7565
Epoch 14/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5199 - accuracy: 0.7713 - val_loss: 0.5154 - val_accuracy: 0.7737
Epoch 15/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5078 - accuracy: 0.7703 - val_loss: 0.5110 - val_accuracy: 0.7715
Epoch 16/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5102 - accuracy: 0.7689 - val_loss: 0.5569 - val_accuracy: 0.7511
Epoch 17/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5082 - accuracy: 0.7700 - val_loss: 0.5086 - val_accuracy: 0.7656
Epoch 18/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4993 - accuracy: 0.7777 - val_loss: 0.5314 - val_accuracy: 0.7597
Epoch 19/40
378/378 [==============================] - 2s 5ms/step - loss: 0.4972 - accuracy: 0.7817 - val_loss: 0.5105 - val_accuracy: 0.7704
Epoch 20/40
378/378 [==============================] - 2s 6ms/step - loss: 0.5016 - accuracy: 0.7801 - val_loss: 0.5017 - val_accuracy: 0.7608
Epoch 21/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4932 - accuracy: 0.7795 - val_loss: 0.5147 - val_accuracy: 0.7704
Epoch 22/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4906 - accuracy: 0.7769 - val_loss: 0.4906 - val_accuracy: 0.7710
Epoch 23/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4828 - accuracy: 0.7782 - val_loss: 0.5322 - val_accuracy: 0.7613
Epoch 24/40
378/378 [==============================] - 2s 5ms/step - loss: 0.5045 - accuracy: 0.7750 - val_loss: 0.4999 - val_accuracy: 0.7758
Epoch 25/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4873 - accuracy: 0.7782 - val_loss: 0.4928 - val_accuracy: 0.7672
Epoch 26/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4802 - accuracy: 0.7864 - val_loss: 0.4907 - val_accuracy: 0.7753
Epoch 27/40
378/378 [==============================] - 2s 5ms/step - loss: 0.4820 - accuracy: 0.7803 - val_loss: 0.4916 - val_accuracy: 0.7694
Epoch 28/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4745 - accuracy: 0.7888 - val_loss: 0.5212 - val_accuracy: 0.7683
Epoch 29/40
378/378 [==============================] - 2s 5ms/step - loss: 0.4761 - accuracy: 0.7809 - val_loss: 0.4911 - val_accuracy: 0.7769
Epoch 30/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4822 - accuracy: 0.7843 - val_loss: 0.4918 - val_accuracy: 0.7758
Epoch 31/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4780 - accuracy: 0.7862 - val_loss: 0.4957 - val_accuracy: 0.7704
Epoch 32/40
378/378 [==============================] - 2s 5ms/step - loss: 0.4761 - accuracy: 0.7870 - val_loss: 0.4826 - val_accuracy: 0.7720
Epoch 33/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4770 - accuracy: 0.7817 - val_loss: 0.4954 - val_accuracy: 0.7769
Epoch 34/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4741 - accuracy: 0.7867 - val_loss: 0.4839 - val_accuracy: 0.7790
Epoch 35/40
378/378 [==============================] - 2s 5ms/step - loss: 0.4737 - accuracy: 0.7859 - val_loss: 0.4925 - val_accuracy: 0.7769
Epoch 36/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4756 - accuracy: 0.7795 - val_loss: 0.5081 - val_accuracy: 0.7753
Epoch 37/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4710 - accuracy: 0.7901 - val_loss: 0.4929 - val_accuracy: 0.7769
Epoch 38/40
378/378 [==============================] - 2s 5ms/step - loss: 0.4756 - accuracy: 0.7843 - val_loss: 0.4877 - val_accuracy: 0.7715
Epoch 39/40
378/378 [==============================] - 2s 6ms/step - loss: 0.4709 - accuracy: 0.7893 - val_loss: 0.5081 - val_accuracy: 0.7763
Epoch 40/40
378/378 [==============================] - 2s 5ms/step - loss: 0.4719 - accuracy: 0.7809 - val_loss: 0.5024 - val_accuracy: 0.7758
In [233]:
fig_3 = go.Figure()
In [234]:
fig_3.add_trace(go.Scatter(x =np.arange(0,len(model_history_2.history['accuracy'])),
                         y = model_history_2.history['val_accuracy'],
                         mode='lines+markers',
                         name='val_accuracy'))
fig_3.add_trace(go.Scatter(x =np.arange(0,len(model_history_2.history['accuracy'])),
                         y = model_history_2.history['accuracy'],
                         mode='lines+markers',
                         name='Accuracy'))
fig_3.update_layout(title = 'ACCURACY vs VALIDATION_ACCURACY')

fig_3.update_xaxes(title_text="Epochs")
fig_3.update_yaxes(title_text="Accuracy")

fig_3.show()
In [236]:
fig_4 = go.Figure()
In [237]:
fig_4.add_trace(go.Scatter(x =np.arange(0,len(model_history_2.history['loss'])),
                         y = model_history_2.history['loss'],
                         mode='lines+markers',
                         name='loss'))
fig_4.add_trace(go.Scatter(x =np.arange(0,len(model_history_2.history['loss'])),
                         y = model_history_2.history['val_loss'],
                         mode='lines+markers',
                         name='val_loss'))
fig_4.update_layout(title = 'LOSS vs VALIDATION_LOSS')

fig_4.update_xaxes(title_text="Epochs")
fig_4.update_yaxes(title_text="Loss")

fig_4.show()
In [217]:
ANN_model_2.evaluate(x_test, y_test)

45/45 [==============================] - 0s 3ms/step - loss: 0.4547 - accuracy: 0.8006
Out[217]:
[0.45469385385513306, 0.8005678057670593]

The Accuracy after removing least importance features is 80%.¶

We improved 2% Accuracy from the previous model¶

In [218]:
predict_2 = ANN_model_2.predict(x_test)

45/45 [==============================] - 0s 1ms/step
In [219]:
predict_2
Out[219]:
array([[0.1171041 ],
       [0.1171041 ],
       [0.1171041 ],
       ...,
       [0.51430327],
       [0.64881563],
       [0.67293304]], dtype=float32)
In [220]:
# Converting those predicted probabilities into Binary output
predict_new_2 = []
for x in predict_2:
    if x >= 0.5:
        predict_new_2.append(1)
    else:
        predict_new_2.append(0)
In [221]:
predict_new_2[-10 : ]
Out[221]:
[0, 0, 0, 0, 0, 0, 0, 1, 1, 1]
In [222]:
plx.imshow(confusion_matrix( y_test, predict_new_2), text_auto = True)
In [223]:
print(classification_report(y_test, predict_new_2))

              precision    recall  f1-score   support

           0       0.82      0.94      0.88      1062
           1       0.67      0.37      0.48       347

    accuracy                           0.80      1409
   macro avg       0.75      0.66      0.68      1409
weighted avg       0.78      0.80      0.78      1409

The Accuracy of 2nd model is 80%¶

In [226]:
import pickle as pkl
In [227]:
with open("Telco_customer_churn_prediction_model.pkl", "wb") as f:
    pkl.dump(ANN_model_2, f)

INFO:tensorflow:Assets written to: ram://0fee0b00-391e-40a8-923a-4b64b302f98b/assets

Our model is stored in pickle file successfully.¶